[SOLVED] 5900x high temps?

[PcBuilder845]

I build a new PC with the Ryzen 5900x. But I feel that the CPU temperature is higher than expected (I'm a little bit paranoid when it comes to temperatures )

My build is:

• CPU - AMD Ryzen 5900x (no overclocking and no PBO)
• CPU Cooler - be quiet! Dark Rock Pro 4 (with supplied thermal paste)
• Motherboard - Asus ROG Strix x570-f gaming
• GPU - Gigabyte GeForce RTX 2070 Windforce x3
• Memory - Corsair Vengeance LPX 64GB (2x32GB) ddr4 3200 (with XMP active)

The idle - CPU temperature is 38-40 C and the CPU Package is 49-51 C. But I usually monitor the CPU Package.
With Cinebench the peak CPU Package temperature is 78 C.
Under heavy gaming, the CPU Package temperature can peak up to 86-90 C. The CPU temperature can peak up to 75 C.

I measured the temperatures with HWMonitor and AI Suite 3.

90 C of CPU Package at peak feels too high for me. Is that normal for 5900x or I have any issues?

 

[CountMike]

I build a new PC with the Ryzen 5900x. But I feel that the CPU temperature is higher than expected (I'm a little bit paranoid when it comes to temperatures )

My build is:

• CPU - AMD Ryzen 5900x (no overclocking and no PBO)
• CPU Cooler - be quiet! Dark Rock Pro 4 (with supplied thermal paste)
• Motherboard - Asus ROG Strix x570-f gaming

The idle - CPU temperature is 38-40 C and the CPU Package is 49-51 C. But I usually monitor the CPU Package.
With Cinebench the peak CPU Package temperature is 78 C.
Under heavy gaming, the CPU Package temperature can peak up to 86-90 C. The CPU temperature can peak up to 75 C.

I measured the temperatures with HWMonitor and AI Suite 3.

90 C of CPU Package at peak feels too high for me. Is that normal for 5900x or I have any issues?

Zen3 (5000 series) are allowed much higher temps than previous Ryzen but yes, 90c is at very limit when it should throttle. I'm afraid that CPU cooler from that segment is just not up to it 100%. Air coolers in general are highly influenced by case cooling but that seems to be overlooked a lot.
On the other hand, performace of Ryzen CPUs is affected mostly by CPU temps and 75c is pretty good.
Apart for cooling, voltage management is most important and various BIOS versions tend to set it on a high side (in the name of stability), introducing negative voltage offset can help a lot.

 

[PcBuilder845]

Zen3 (5000 series) are allowed much higher temps than previous Ryzen but yes, 90c is at very limit when it should throttle. I'm afraid that CPU cooler from that segment is just not up to it 100%. Air coolers in general are highly influenced by case cooling but that seems to be overlooked a lot.
On the other hand, performace of Ryzen CPUs is affected mostly by CPU temps and 75c is pretty good.
Apart for cooling, voltage management is most important and various BIOS versions tend to set it on a high side (in the name of stability), introducing negative voltage offset can help a lot.

Thank you for the reply!

To be honest, my latest PC build was 5-6 years ago and it was actually much easier 😀. I'm so confused about how to monitor these Ryzen temps... There are CPU temperature, CPU Package, CPU CCD1, CPU temperature in the Ryzen master (which is different from the rest - I didn't find which value its monitor). I know that the Motherboard monitors the CPU temperature to control the fans. And actually, I'm not sure which temperature should I monitor and keep under Ryzen 90 C limit...
I tried to look over the internet and I feel that people often talking about different temperatures. For example, I saw a thread that someone said his "CPU temperature" and it was similar to my CPU Temperature but someone sent a comment and said that his "CPU Temperature" is higher, but it was similar to my CPU Package temperature... So I believe that it is not only me who confused about which temperature we should maintain cool (Ok, all of them... ), but which one should be kept under 90 C?

About the PC Case. The PC Case that I have is the Antec P110 Silent with 3 front fans and 1 rear fan. Each one is 120mm and working constantly on 100%. Although I have an option to add 2 top fans, I found it inserts a lot of dust when the PC turned off and releases a lot of noise when it turned on. Two fans in the case (1 front and 1 rear) are the stock fans that came with the case, and the other two front fans are cheap ones (that I already had). When I put my hand in the case, I can definitely feel a good wind from the front fans (although it seems that the front panel is solid close).
Could better case fans reduce the CPU temperature?

In the Cinebench the temperatures are normals. Maybe the heat of the other components while gaming (GPU mostly?) raises up to the CPU and the airflow is not enough to handle it...
Anyway, I will try to monitor all the CPU temperatures (that I mentioned above) while gaming and post the screenshots here.

About the voltage, I've never done over- or underclocking. And since it is a pretty new build (two days old ) I prefer not to avoid the warranty at this point or break something .

 

[CountMike]

Thank you for the reply!

To be honest, my latest PC build was 5-6 years ago and it was actually much easier 😀. I'm so confused about how to monitor these Ryzen temps... There are CPU temperature, CPU Package, CPU CCD1, CPU temperature in the Ryzen master (which is different from the rest - I didn't find which value its monitor). I know that the Motherboard monitors the CPU temperature to control the fans. And actually, I'm not sure which temperature should I monitor and keep under Ryzen 90 C limit...
I tried to look over the internet and I feel that people often talking about different temperatures. For example, I saw a thread that someone said his "CPU temperature" and it was similar to my CPU Temperature but someone sent a comment and said that his "CPU Temperature" is higher, but it was similar to my CPU Package temperature... So I believe that it is not only me who confused about which temperature we should maintain cool (Ok, all of them... ), but which one should be kept under 90 C?

About the PC Case. The PC Case that I have is the Antec P110 Silent with 3 front fans and 1 rear fan. Each one is 120mm and working constantly on 100%. Although I have an option to add 2 top fans, I found it inserts a lot of dust when the PC turned off and releases a lot of noise when it turned on. Two fans in the case (1 front and 1 rear) are the stock fans that came with the case, and the other two front fans are cheap ones (that I already had). When I put my hand in the case, I can definitely feel a good wind from the front fans (although it seems that the front panel is solid close).
Could better case fans reduce the CPU temperature even more?

About the voltage, I've never done over- or underclocking. And since it is a pretty new build (two days old ) I prefer not to avoid the warranty at this point or break something .

In general, trust what Ryzen Master tells you, it shows CORE temps and only for highest core temperature.
Setting negative voltage temp in BiOS is painless but needs some experimenting, -00.5 to - 0.1v is usual amount and in no way would influence warranty.

 

[Phaaze88]

When I put my hand in the case, I can definitely feel a good wind from the front fans (although it seems that the front panel is solid close).
Could better case fans reduce the CPU temperature?

1)Does not count, because by opening the chassis to feel the airflow, you completely changed the air pressure.
2)No... well, with the kind of money you'd have to spend for 'em, and the higher noise volume that'll come with it, I'm sure you won't care for it.

I don't see any mention of what the gpu is...

 

[PcBuilder845]

1)Does not count, because by opening the chassis to feel the airflow, you completely changed the air pressure.
2)No... well, with the kind of money you'd have to spend for 'em, and the higher noise volume that'll come with it, I'm sure you won't care for it.

I don't see any mention of what the gpu is...

Thanks for the reply!

GPU: Gigabyte GeForce RTX 2070 Windforce x3
I will add this to the original post

 

[drea.drechsler]

Thank you for the reply!

To be honest, my latest PC build was 5-6 years ago and it was actually much easier 😀. I'm so confused about how to monitor these Ryzen temps... There are CPU temperature, CPU Package, CPU CCD1, CPU temperature in the Ryzen master (which is different from the rest - I didn't find which value its monitor). I know that the Motherboard monitors the CPU temperature to control the fans. And actually, I'm not sure which temperature should I monitor and keep under Ryzen 90 C limit...
I tried to look over the internet and I feel that people often talking about different temperatures. For example, I saw a thread that someone said his "CPU temperature" and it was similar to my CPU Temperature but someone sent a comment and said that his "CPU Temperature" is higher, but it was similar to my CPU Package temperature... So I believe that it is not only me who confused about which temperature we should maintain cool (Ok, all of them... ), but which one should be kept under 90 C?

About the PC Case. The PC Case that I have is the Antec P110 Silent with 3 front fans and 1 rear fan. Each one is 120mm and working constantly on 100%. Although I have an option to add 2 top fans, I found it inserts a lot of dust when the PC turned off and releases a lot of noise when it turned on. Two fans in the case (1 front and 1 rear) are the stock fans that came with the case, and the other two front fans are cheap ones (that I already had). When I put my hand in the case, I can definitely feel a good wind from the front fans (although it seems that the front panel is solid close).
Could better case fans reduce the CPU temperature?

In the Cinebench the temperatures are normals. Maybe the heat of the other components while gaming (GPU mostly?) raises up to the CPU and the airflow is not enough to handle it...
Anyway, I will try to monitor all the CPU temperatures (that I mentioned above) while gaming and post the screenshots here.

About the voltage, I've never done over- or underclocking. And since it is a pretty new build (two days old ) I prefer not to avoid the warranty at this point or break something .

Welcome to the modern world of highly dynamic CPU's that boost individual cores aggressively even at idle.

What are you using to read temperature? Ryzenmaster's good...but it's pretty single minded as it reports an average core temp only. Do not use HWMonitor or Aida64 or any motherboard utility as they can be pretty inaccurate at times.

Get HWInfo64 and use it. HWInfo has both die temp readings and an average core temp, it's good to know both since the CPU boosts so aggressively it can spike temps along with the boost. A 5900X can have literally hundreds of temp sensors scattered all around the CPU cores, so the spike is just the hottest of one of those sensors when it boosts. It's the average temp that's important as it's closer to the true thermal state of the CPU.

There can be other readings (like die temp) since the CPU has multiple chiplets as well as an I/O chiplet. People often overclock just one chiplet so knowing it's temp is helpful.

Package Temp can depend on your motherboard...if it has a temp sensor in the area of the package for instance. It's not very useful, IMO, as it will lag performance surges (did I not say Ryzen is extremely dynamic?) by a lot.

90C is Tjmax for 5900X. If you're hitting that with an average temp and BIOS is setup stock then something's not good about your cooling mounting....if only while gaming check that your cooler isn't just sucking in the GPU's hot output air. Spikes to 90C (use HWInfo64, right click on the sensor reading and Show Graph to see it move across time) might be OK, but I'd expect a DR-pro-4 to be much better than that.

 

[PcBuilder845]

Welcome to the modern world of highly dynamic CPU's that boost individual cores aggressively even at idle.

What are you using to read temperature? Ryzenmaster's good...but it's pretty single minded as it reports an average core temp only. Do not use HWMonitor or Aida64 or any motherboard utility as they can be pretty inaccurate at times.

Get HWInfo64 and use it. HWInfo has both die temp readings and an average core temp, it's good to know both since the CPU boosts so aggressively it can spike temps along with the boost. A 5900X can have literally hundreds of temp sensors scattered all around the CPU cores, so the spike is just the hottest of one of those sensors when it boosts. It's the average temp that's important as it's closer to the true thermal state of the CPU.

There can be other readings (like die temp) since the CPU has multiple chiplets as well as an I/O chiplet. People often overclock just one chiplet so knowing it's temp is helpful.

Package Temp can depend on your motherboard...if it has a temp sensor in the area of the package for instance. It's not very useful, IMO, as it will lag performance surges (did I not say Ryzen is extremely dynamic?) by a lot.

90C is Tjmax for 5900X. If you're hitting that with an average temp and BIOS is setup stock then something's not good about your cooling mounting. Spikes to 90C (use HWInfo64, right click on the sensor reading and Show Graph to see it move across time) might be OK, but I'd expect a DR-pro-4 to be much better than that.

Thank you for the reply!

I've monitored the CPU Package with HWMonitor when I saw 90 C. Maybe it was incorrect since the AI Suite 3 (Fan Xpert) set CPU fans RPM as the CPU temperature was 75 C. I was surprised either since I heard that DRP4 is one of the best coolers for the x5900... I didn't plan to overclock it (not yet for sure) but it seems to be impossible with these resutes...

I will run a game and Cinebench later while monitoring the CPU with different tools and share the results...

 

[drea.drechsler]

Seriously

Thank you for the reply!

I've monitored the CPU Package with HWMonitor when I saw 90 C.
...

Seriously, get HWinfo64 and stop guessing. You'll have consistent and logical readings that can make sense...and an explanation of the sensor reading when you hover your mouse over it so you can understand what makes it differ from another that's similar. I wouldn't attempt overclocking a 5900X unless you have it.

Actually, I wouldn't attempt overclocking a 5900X unti you understand it's operation and boosting. The boost algorithm works so well that all-core overclocking Ryzen is rarely very fruitful (especially Zen 3) and can take a lot of tweaking and extremely good cooling. Even then it might only benefit heavy, all-core workload performance and actually hurt it for lightly threaded workloads such as gaming.

 

[PcBuilder845]

Zen3 (5000 series) are allowed much higher temps than previous Ryzen but yes, 90c is at very limit when it should throttle. I'm afraid that CPU cooler from that segment is just not up to it 100%. Air coolers in general are highly influenced by case cooling but that seems to be overlooked a lot.
On the other hand, performace of Ryzen CPUs is affected mostly by CPU temps and 75c is pretty good.
Apart for cooling, voltage management is most important and various BIOS versions tend to set it on a high side (in the name of stability), introducing negative voltage offset can help a lot.

Seriously

Seriously, get HWinfo64 and stop guessing. You'll have consistent and logical readings that can make sense...and an explanation of the sensor reading when you hover your mouse over it so you can understand what makes it differ from another that's similar. I wouldn't attempt overclocking a 5900X unless you have it.

Actually, I wouldn't attempt overclocking a 5900X unti you understand it's operation and boosting. The boost algorithm works so well that all-core overclocking Ryzen is rarely very fruitful (especially Zen 3) and can take a lot of tweaking and extremely good cooling. Even then it might only benefit heavy, all-core workload performance and actually hurt it for lightly threaded workloads such as gaming.

Here are some tests that I run (yea, I understand that HWinfo64 is the best, however, I monitored with all my monitoring apps )

Cinebench results


These results look normal, although the CPU CCD1 reached 89 C at the peak... This one spike can be seen in the graph. The funniest thing is that my score was slightly lower, probably since all the monitoring apps in the background .

While Gaming


These although look little bit higher than Cinebench. The CPU CCD1, Die and Package reached ~90 C at the peak... But the graphs are pretty stable at 80 C (I run the game for a while, and I didn't see the peak in the graphs, maybe it was earlier)...
The motherboard reads 72 C (?) for controlling the CPU fans (I set that >70C should run at 100%). Ryzen Master shows 82.5C and I tried to constantly check it (since it is not capture the min/max) but all the time I checked it was around ~80 C. I didn't see it reach near to 90 C (but since I manually tested it, it is easy to miss)... The graphs and the average temps in HWinfo64 seems to be stable around 80 C so I believe that 90 C are peaks.

I'm just worried if these 90 C peaks are important or it normal behavior for that processor... Any info will be highly appreciated!

 

[drea.drechsler]

Here are some tests that I run (yea, I understand that HWinfo64 is the best, however, I monitored with all my monitoring apps )

Cinebench results


These results look normal, although the CPU CCD1 reached 89 C at the peak... This one spike can be seen in the graph. The funniest thing is that my score was slightly lower, probably since all the monitoring apps in the background .

While Gaming


These although look little bit higher than Cinebench. The CPU CCD1, Die and Package reached ~90 C at the peak... But the graphs are pretty stable at 80 C (I run the game for a while, and I didn't see the peak in the graphs, maybe it was earlier)...
The motherboard reads 72 C (?) for controlling the CPU fans (I set that >70C should run at 100%). Ryzen Master shows 82.5C and I tried to constantly check it (since it is not capture the min/max) but all the time I checked it was around ~80 C. I didn't see it reach near to 90 C (but since I manually tested it, it is easy to miss)... The graphs and the average temps in HWinfo64 seems to be stable around 80 C so I believe that 90 C are peaks.

I'm just worried if these 90 C peaks are important or it normal behavior for that processor... Any info will be highly appreciated!

I'm not that familiar with Ryzen 5000...but I'm aware they are pretty hot blooded. I think the 90C peaks are localized spikes and not thermally significant; note that they are only seen in CCD1 in HWInfo (actually 89C but close enough) reading. I wonder if your gold star cores are in that CCD?

But what IS thermally significant is the CPU die (average) reading which never exceeds 82C. That's a running average of all the sensors across both CPU die. 82C for 12 cores in an AVX heavy workload like CB23 doesn't seem very bad to me. it's also very consistent with Ryzenmaster (the current reading) which is a good averaging temp reading and also HWMonitor.

What would be interesting is to put each of the CCD1 and CCD0 sensors, Tdie/Tctl, and CPU Die (average) readings in a graph (right click on the sensor reading, show graph) and watch them through time from idle to ramping up in CB23 then as it ramps down after. I bet you'll see almost a saw-tooth type chart of Tdie/Tctl and the CCDn sensors while the average floats up and then back down.

Missed it...you did! so the sawtooth are the little spikes as one core saw an opportunity with power, temperature and the right processing load to boost just a bit higher and so the temp spikes briefly. It's the algorithm doing it's thing to deliver maximum performance whenever it can.

 

[PcBuilder845]

I'm not that familiar with Ryzen 5000...but I'm aware they are pretty hot blooded. I think the 90C peaks are localized spikes and not thermally significant; note that they are only seen in CCD1 in HWInfo (actually 89C but close enough) reading. I wonder if your gold star cores are in that CCD?

But what IS thermally significant is the CPU die (average) reading which never exceeds 82C. That's a running average of all the sensors across both CPU die. 82C for 12 cores in an AVX heavy workload like CB23 doesn't seem very bad to me. it's also very consistent with Ryzenmaster (the current reading) which is a good averaging temp reading and also HWMonitor.

What would be interesting is to put each of the CCD1 and CCD0 sensors, Tdie/Tctl, and CPU Die (average) readings in a graph (right click on the sensor reading, show graph) and watch them through time from idle to ramping up in CB23 then as it ramps down after. I bet you'll see almost a saw-tooth type chart of Tdie/Tctl and the CCDn sensors while the average floats up and then back down.

Missed it...you did! so the sawtooth are the little spikes as one core saw an opportunity with power, temperature and the right processing load to boost just a bit higher and so the temp spikes briefly. It's the algorithm doing it's thing to deliver maximum performance whenever it can.

It seems that the CPU Die (average) has reached 90C is some peaks while gaming (second image). Although the average value while gaming is 78.8C, so I believe it is fine(?...)

If Cinebench R23 creates a heavy CPU load (CPU Peak speed is <4.0GHz??) and reach up to 82C in CPU Die (average), without using GPU (correct me if I'm wrong), maybe the CPU cooler is fine. But with gaming and GPU running at full power can also create heat that raises up to the CPU and maybe harm the CPU cooler efficiency. Could better / more case fans help to reduce the CPU temperature while gaming?

 

[drea.drechsler]

... But with gaming and GPU running at full power can also create heat that raises up to the CPU and maybe harm the CPU cooler efficiency. Could better / more case fans help to reduce the CPU temperature while gaming?

I would agree with that...and I think that's what @Phaaze88 was driving at earlier. Now that you know what you're seeing with temps, you can better assess cooling performance in various workloads.

Air cooling is very dependent on good case ventilation. The Antec P110 doesn't look very good in that regard...it has 3 front fans but they are have very restricted air flow by the front panel. Then only one case fan to pull the hot air back out.

 

[Phaaze88]

Any cooling - and I mean ANY - is only as effective/efficient as what the user's chassis allows.

Spoiler

Some users want silent fans.
Manufactures are like, "Low rpm fans, huh? We can do that. Cooling might take a hit."

They want dust filtration everywhere, or a dust-free(nigh impossible) PC.
"Ok. Cooling takes a tiny bump, but this easily doable."

They want to hide the cable mess from the power supply.
"Check! Power supply shrouds are available! But cooling takes another bump."

They want soundproofing paneling everywhere.
"Ok? That doesn't really work on the high frequencies produced by fans, but - done! Cooling again though..."

They want RGB fans.
"Check. We had to lower the power profile a little because of the LEDs, so they're slightly weaker than the original."

They want to be able to cram high power consuming cpus and gpus, and for the cooling to be good.
"...
%*#&!"

All of those desires may not apply to you, but they all add up.

The DRP4 is a good cooler. The Windforce 3x cooler on the gpu is also good.
But if air doesn't get in and out effectively, what happens?

Hybrid coolers aren't immune to this either - they too, need air. They'll get just as hot, but it'll take longer.
While you could brute force it on the liquid alone, the cooler won't last as long because the parts inside the pump are being exposed to higher fluid temps than normal; they break down faster.

The PC is also a space heater. It doesn't benefit it or the user for warm air to get trapped/cycled inside, or for it to take more time to escape its container.

 

[PcBuilder845]

That's true, and maybe I need to replace or add fans. I was afraid of overheating the CPU, but now I have more confidence that it is in the normal curve (filtering the peaks from the graph....). My two cheaps fans have a Molex connector, so I don't know if they are still running at the desired RPM (but the good this is that they run at 100% all the time )... I wish I had a way to measure the airflow inside the case when it closed The only indicator when the case is closed is the airflow on the rear side and I can feel the heat came out (with a good airflow)

And a comment about your spoiler. I do have a problem with noise from the PC, it can make me crazy (the rest are less important to me). But it is the first time I bought be quiet! product and it is an excellent cooler and super quiet. I can't hear it even if it runs on 100%. So maybe there are some fan manufacturers that can do both .

 

[drea.drechsler]

So maybe there are some fan manufacturers that can do both .

Noctua's got the last word in 'best fan'. But here it is along with some more of the best fan choices being tested head to head...

View: https://www.youtube.com/watch?v=QwftVMGPOiI


It doesn't have to be really expensive either. Arctic Cooler P12's you can get a 5 pack for around $40 US. He reviews that against the Noctua too.

 

[PcBuilder845]

BTW, would you recommend undervolting or these temps are good as now?

 

[drea.drechsler]

BTW, would you recommend undervolting or these temps are good as now?

Undervolting is a definite possibility with 5900X's...there's a tool called the 'curve optimizer', a part of 'pbo2', that benefits from it. I've no idea how to do it though. I'd google it and start reading up on it as it sounds like some people have had success and its a good way to go.

One thing I've learned with my 3700X: undervolting can be misleading. You think it's going just fine but fail to realize it's hurting light threaded performance. In such cases, go very slightly and only use offset adjusments. Test with Cinebench23 single thread or two thread benchmark runs to see if it's hurting or at least staying unchanged then try a little more.